[GitHub] szha closed pull request #8995: Fix race condition in engine start/stop

[GitBox <gi...@apache.org>]

szha closed pull request #8995: Fix race condition in engine start/stop
URL: https://github.com/apache/incubator-mxnet/pull/8995
 
 
   

This is a PR merged from a forked repository.
As GitHub hides the original diff on merge, it is displayed below for
the sake of provenance:

As this is a foreign pull request (from a fork), the diff is supplied
below (as it won't show otherwise due to GitHub magic):

diff --git a/src/engine/threaded_engine_perdevice.cc b/src/engine/threaded_engine_perdevice.cc
index 28bc92f7b2..e7e222f6cb 100644
--- a/src/engine/threaded_engine_perdevice.cc
+++ b/src/engine/threaded_engine_perdevice.cc
@@ -95,9 +95,10 @@ class ThreadedEnginePerDevice : public ThreadedEngine {
     int cpu_priority_nthreads = dmlc::GetEnv("MXNET_CPU_PRIORITY_NTHREADS", 4);
     cpu_priority_worker_.reset(new ThreadWorkerBlock<kPriorityQueue>());
     cpu_priority_worker_->pool.reset(new ThreadPool(
-        cpu_priority_nthreads, [this]() {
-          this->CPUWorker(Context(), cpu_priority_worker_.get());
-        }));
+        cpu_priority_nthreads,
+        [this](std::shared_ptr<ThreadPool::SimpleEvent> ready_event) {
+          this->CPUWorker(Context(), cpu_priority_worker_.get(), ready_event);
+        }, true));
     // GPU tasks will be created lazily
   }
 
@@ -122,9 +123,10 @@ class ThreadedEnginePerDevice : public ThreadedEngine {
           auto ptr =
           cpu_normal_workers_.Get(dev_id, [this, ctx, nthread]() {
               auto blk = new ThreadWorkerBlock<kWorkerQueue>();
-              blk->pool.reset(new ThreadPool(nthread, [this, ctx, blk] () {
-                    this->CPUWorker(ctx, blk);
-                  }));
+              blk->pool.reset(new ThreadPool(nthread,
+                  [this, ctx, blk](std::shared_ptr<ThreadPool::SimpleEvent> ready_event) {
+                    this->CPUWorker(ctx, blk, ready_event);
+                  }, true));
               return blk;
             });
           if (ptr) {
@@ -259,12 +261,14 @@ class ThreadedEnginePerDevice : public ThreadedEngine {
    */
   template<dmlc::ConcurrentQueueType type>
   inline void CPUWorker(Context ctx,
-                        ThreadWorkerBlock<type> *block) {
+                        ThreadWorkerBlock<type> *block,
+                        std::shared_ptr<ThreadPool::SimpleEvent> ready_event) {
     this->is_worker_ = true;
     auto* task_queue = &(block->task_queue);
     RunContext run_ctx{ctx, nullptr};
     // execute task
     OprBlock* opr_block;
+    ready_event->signal();
     while (task_queue->Pop(&opr_block)) {
       this->ExecuteOprBlock(run_ctx, opr_block);
     }


 

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services